Apple Music’s ‘New Music Mix’ Is Confusing Today for Some People ⇥ reddit.com

My listening habits on Apple Music this week have largely been a blend of shoegaze, post rock, folk, and punk — much like last week; maybe this is a zeitgeist thing — so you can imagine my surprise as I opened my New Music Mix this morning to find the following:

• a beat poem by Soundwalk Collective,

• a collaboration between the Black Eyed Peas and an Israeli dance pop duo,

• an Indian pop song,

• and then twenty-two psy-trance tracks in a row.

I am used to a solid New Music Mix and the occasional oddball selection, but this is so far off from what I listen to that it seems some machine learning variable used the wrong operation and gave me the least-relevant playlist possible.

This is not something only I am seeing. There are a few discussions on Reddit and the MacRumors forums with complaints of inexplicable music choices. I asked on Twitter and in Slack and a few others reported the same thing.

I would wager that this is not happening to most people. There is clearly self-selection bias, too. But there seem to be some patterns in what some people are seeing in their New Music Mix this week: lots of classical music, lots of dance music, and some Indian and Spanish pop songs.

As a writer of a tech-focused website, I would love to know how something like this happens from a machine learning standpoint.¹ But, as a user, I just want a button stronger than “dislike” — some option that lets me mark a generated playlist as entirely backwards and wrong for me.